Author: Ian Osgood
Date: 15:43:55 12/19/02
Go up one level in this thread
On December 18, 2002 at 21:14:48, Ian Osgood wrote: >Going from one to two cpus sped up bench from 267 seconds to 175 seconds for >about 48M nodes (which varies from 41M to 62M due to multi-processing >indeterminacy). Actually, the indeterminacy is due to bugs in the hash table: looking back at the logs, I am seeing many instances of "bad move from hash table" (from ply 6 onwards) and "captured a king". The first game played was flawless, but most other tests and games have numerous hash errors (once aborting with two pieces on the same square). This happens even with no optimization and SMP undefined, so it is a basic problem on this platform. Ian Osgood
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.